In recent years, applications for user authentication by an ID card such as a driver's license or a basic resident registration card have been widespread. For such authentication, a method is typically carried out such that an ID card is scanned to acquire its digital image, the information on the front face thereof is automatically read by image recognition, and based on the information a user is authenticated. Also, cards with increased security with recorded fine patterns or characters on the card surface have become widespread in recent years, and an image scanner product that is equipped with a function to automatically recognize such a card and compare it to a high resolution color image have been introduced.
In such an image recognition system, a medium is divided into partial portions for multiple scanning in order to perform image recognition on a medium which has a large capacity exceeding the ability of the imaging device or [the capacity of] the image memory, and then those partial images are combined with each other afterwards to re-produce the whole image. As an example of such a technique for combining multiple [partial] images, Patent reference 1 has proposed a technique in which a spatial frequency is acquired by using discrete cosine transform (DCT) over the combining portions of the images, and the width of the combining portion is varied according to the size of the spatial frequency to implement the image-combining with fewer errors.